|
|
--- |
|
|
base_model: |
|
|
- unsloth/Llama-3.2-1B |
|
|
tags: |
|
|
- text-generation-inference |
|
|
- transformers |
|
|
- math |
|
|
- conversational |
|
|
- llama |
|
|
- meta |
|
|
license: apache-2.0 |
|
|
language: |
|
|
- en |
|
|
library_name: transformers |
|
|
--- |
|
|
# GsMath-Llama-1B |
|
|
## Model Description: |
|
|
This is a fine-tuned version of [unsloth/Llama-3.2-1B](https://huggingface.co/unsloth/Llama-3.2-1B)! |
|
|
- **recommended settings for inference:** min_p = 0.1 and temperature = 1.5 , Read this [Tweet](https://x.com/menhguin/status/1826132708508213629) to understand why. |
|
|
- **License :** apache-2.0 |
|
|
- **Finetuned from model :** unsloth/Llama-3.2-1B |
|
|
## Benchmarks: |
|
|
We evaluate both models on GSM8K using the standard lm-eval 5-shot exact-match protocol. Under identical decoding and extraction settings,GsMath-Llama-1B outperforms Meta’s Llama-3.2-1B by 2x,demonstrating an improvement in small-model mathematical capability. |
|
|
| Model | Params | GSM8K (5-shot, EM) | |
|
|
| ----------------------------- | ------ | ------------------ | |
|
|
| **GsMath-Llama-1B** | 1B | **0.137** | |
|
|
| Llama-3.2-1B | 1B | 0.068 | |
|
|
|
|
|
|
|
|
<p align="left"> |
|
|
<img alt="GsMath-Llama-1B" src="https://huggingface.co/CannaeAI/GRPO-Gsmath-Llama-1B-IT/resolve/main/GSM8K-IT.png"> |
|
|
</p> |
|
|
|
